augusto kiniama rosa

Posted on Oct 21 • Originally published at blog.archetypeconsulting.com on Oct 2

The Unofficial Snowflake Monthly Release Notes: September 2025

#ai #snowflake #data #newreleases

Monthly Snowflake Unofficial Release Notes #New features #Previews #Clients #Behavior Changes

Welcome to the September 2025 Unofficial Release Notes for Snowflake! Here, you’ll find all the latest features, drivers, and more in one convenient place.

As an unofficial source, I am excited to share my insights and thoughts. Let’s dive in! You can also find all of Snowflake’s releases here.

This month, we provide coverage up to release 9.29 (General Availability — GA). I hope to extend this eventually to private preview notices as well.

I would appreciate your suggestions on continuing to combine these monthly release notes. Feel free to comment below or chat with me on LinkedIn.

Behavior change bundle 2025_04 is generally enabled for all customers, 2025_05 is enabled by default but can be opted out until next BCR deployment, and 2025_06 is disabled by default but may be opted in.

What’s New in Snowflake

New Features

Snowpipe Streaming with high-performance architecture (GA), new architecture is designed from the ground up for large-scale, real-time data ingestion with high throughput and low latency. Key features and benefits: Ingest data at up to 10 GB per second per table, with ingest-to-query latencies typically under 10 seconds, multi-language client support: Java and Python SDKs, built on a shared Rust core for efficiency, centralize your ingestion logic using the PIPE object, benefit from a new throughput-based pricing model that provides predictable costs
Cost management — Updating budgets more frequently, the time period between consumption and a budget receiving information about the consumption is called the budget refresh interval, as short as an hour
FILE data type (GA), enables multimodal AISQL workflows with unstructured data stored on internal or external stages. FILE values provide a way to reference files without encapsulating the actual file content. FILE objects let you: Store references to files in tables and pass them to AISQL functions, avoid duplicating file data and process it more efficiently by creating ad passing FILE values as references, integrate with existing data architectures by combining DIRECTORY functions with TO_FILE

Snowsight Updates

Workspaces (GA), a unified editor for creating and managing code across file types. Workspaces enhance SQL editing with nested folders, rich editing, Copilot help, better charting, and column stats. All workspace content is file-based, simplifying complex projects and Git integration for version control, collaboration, and workflow consistency
Query insights in Snowsight (Preview), the Query Profile tab under Query History now displays insights about conditions that affect query performance that explains how query performance might be affected and provides a general recommendation for next steps

AI Updates (Cortex, ML, DocumentAI)

Document AI models in the model registry, Now stores published or trained models in the Snowflake Model Registry, allowing copying between databases or schemas within or across accounts for easy management, versioning, and RBAC. The registry acts as a control plane for deploying Document AI model versions safely and efficiently across environments
Cortex Agents: Admin object REST API (Preview), Create an agent via REST API and integrate it into your app to perform tasks or respond to queries. Configure a thread to maintain in-memory context, eliminating the need to send context repeatedly. Includes expanded agent features, Cortex Agents workflow updates, and new REST API endpoints
Support for Snowflake Cortex AISQL in incremental dynamic table refresh, the SELECT clause for dynamic tables in incremental refresh mode. The same availability restrictions as described in AISQL functions apply, and add AI-powered insights directly to your dynamic tables, automatically analyzing data as it updates
AI_FILTER Performance Optimization (Preview), delivers a 2–10x speedup and reduces token usage by up to 60% for suitable queries using AI functions in SELECT, WHERE, and JOIN … ON clauses
Cortex AISQL AI_TRANSLATE (GA), provides industry-leading quality for translations of call transcripts, product reviews, social media comments, and other text content with 20% less token since preview
Page filtering for AI_PARSE_DOCUMENT, includes page filtering capabilities, allowing you to parse specific pages or ranges within large documents, process only the content you need
AI_COUNT_TOKENS AISQL function (Preview), helps you size your AI workloads by calculating the total number of input tokens processed by AISQL large language models and task-specific functions, so you can size queries appropriately before hitting model limits and accurately estimate costs based on input token usage
Cortex Analyst semantic model improvements, support for private facts and metrics, and derived metrics calculated from other metrics using arithmetic operations and functions
Cortex Agents integration for Microsoft Teams and Copilot (Preview), enhances Microsoft Teams and Copilot 365 with multi-agent support, higher-quality responses, multi-turn conversations, and a streamlined setup experience

Snowflake Applications (Container Services, Notebooks, Streamlit, and Applications, Snowconvert, Openflow)

Snowpark Container Services in Google Cloud (GA)
GRANT OWNERSHIP ON NOTEBOOK (GA)
Automated granting of privileges (GA), Snowflake Native App providers can automate privilege granting to add needed privileges to the manifest. This enables the app to create objects in the consumer account without manual privilege grants or object creation by the consumer.
App specifications (GA), Allow a Snowflake Native App provider to specify connection info the app requests. When the consumer installs it, they review and approve or decline
Feature policies (GA), Snowflake Native App administrators can set feature policies to restrict objects that an app can create in the consumer account. They can review the required privileges in the listing before installing
SnowConvert AI Verification (Preview), strengthens SnowConvert AI by automating functional validation of converted database code. AI Verification uses synthetic data generation, AI-driven unit testing, and AI-driven error resolution
Snowflake Native Apps support for FedRAMP on AWS for apps with containers (Preview), apps with containers can be distributed to any Snowflake customer who can use them in FedRAMP region.
Snowflake Openflow — Snowflake Deployments (Preview), run on Snowpark Container Services (SPCS) and provide a streamlined and integrated solution for data integration and connectivity across interoperable storage like Iceberg and Snowflake native storage
Support for Streamlit in Snowflake in the People’s Republic of China (Preview)
Snowconvert — IBM DB2 SQL Support
Snowconvert — Lots of fixes for teradata, MS SQL Server, Oracle, SSIS, Bigquery
Snowconvert — AI Verification step for SQL Server migrations
Snowconvert — Tableau support
Snowconvert — ETL & SSIS
Snowconvert — dbt
Snowconvert — PowerBI Support

Data Lake Updates

Partitioned writes for Apache Iceberg™ tables (Preview), improves compatibility with the wider Iceberg ecosystem and enables faster read queries from external tools. You can now use Snowflake to create and write to both managed and external Iceberg tables with partitioning
Support for position row-level deletes when writing to externally managed Apache Iceberg™ tables or catalog-linked databases on Amazon S3 or Google Cloud (Preview), these deletes are supported when Snowflake performs update, delete, and merge operations on the tables
Support for position row-level deletes when writing to externally managed Apache Iceberg™ tables or catalog-linked databases on Azure (Preview), these deletes are supported when Snowflake performs update, delete, and merge operations on the table files. This feature is a performance improvement for these operations
Prevent data compaction on Snowflake-managed Apache Iceberg™ tables, new ENABLE_DATA_COMPACTION parameter to specify whether Snowflake should perform data compaction on Snowflake-managed Apache Iceberg™ tables. Snowflake still performs compaction on these tables by default
New system function to replace the catalog integration for an externally managed Apache Iceberg™ table, new SYSTEM$SET_CATALOG_INTEGRATION system function to replace the catalog integration associated with an externally managed Apache Iceberg™ table and you can use this function to access the latest Iceberg features for your tables, such as write support for externally managed Iceberg tables

Realtime Data (Hybrid tables & SnowPostGres)

Hybrid table support for Microsoft Azure (Preview)

Security, Privacy & Governance Updates

Classifying views automatically (GA), Set up sensitive data classification so database views get automatically classified at regular intervals. Before, only tables could be automatically classified
Excluding objects from automatic classification (Preview), automatically classifies all sensitive data in a database with a classification profile. You can configure Snowflake to exclude schemas, tables, or columns from being classified, so they are skipped during the process
Using Snowsight to monitor data quality (Preview), the new Data Quality tab is available when you are viewing a table or view in Snowsight, use data profiling to view statistics about your object, such as row counts, null values, and value distributions and monitor the results of data metric functions (DMFs) associated with the object
Multi-factor authentication — Support for one-time passcodes, generate one-time passcodes (OTPs) that users can use as their second factor of authentication when signing in to Snowflake with multi-factor authentication (MFA) and used to provide break glass access
Data lineage for tasks, Determine that data moved from a source to a downstream object due to a task. Selecting the arrow between them shows task information
Using SQL for Cortex Powered Object Descriptions (GA), call a stored procedure, AI_GENERATE_TABLE_DESC, to programmatically generate Cortex Powered Object Descriptions
External OAuth support for Snowflake Open Catalog catalog integration (GA), support External OAuth. To configure a catalog integration with External OAuth, first configure External OAuth in Open Catalog, and then use the new OAUTH_TOKEN_URI parameter for the integration

SQL, Extensibility & Performance Updates

New SYS_CONTEXT function for getting context about applications, sessions, and organizations and the new SYS_CONTEXT function to get context information about: current application, current environment, current session
Read consistency mode for sessions with near-concurrent changes, set the READ_CONSISTENCY_MODE account-level parameter to define the level of consistency guarantees that are required for sessions with near-concurrent changes
Retrieve bind variable values (Preview), retrieve the values of the bind variables for queries that have been executed by using the BIND_VALUES table function in the INFORMATION_SCHEMA schema
You can use the RESAMPLE clause and a set of interpolation functions to fill gaps in time-series data, simplifies the process of generating continuous, uniformly-sampled time-series data
More efficient workload distribution, improves query execution time by detecting and adaptively redistributing workloads across nodes in the warehouse, without user intervention

Data Clean Rooms Updates

Version 10.2, bring the welcome modal: Fixed welcome modal links and UI display issues, and add data step: When no data is available, users can still refresh the tables list in the Add Data step, in case data has recently become available

Marketplace, Listings & Data Sharing

Billing views for Snowflake resellers and distributors, Views in the new BILLING schema offer billing data for Snowflake resellers' customers. For example, PARTNER_CONTRACT_ITEMS reveals reseller-customer contracts. Only resellers and distributors can access these views
Declarative Sharing (Preview), allows providers to share and sell data products, enhanced by Snowflake Notebooks to help Snowflake consumers visualize and explore the data. Declarative Sharing’s simplified development experience makes it easy to get started quickly
Declarative Shared Native Apps (Preview), allows providers to share and sell data products, enhanced by Snowflake Notebooks to help Snowflake consumers visualize and explore the data and simplifies development experience makes it easy to get started quickly

Open-Source Updates

terraform-snowflake-provider 2.7.0 (support for generation 2 Standard warehouses and resource constraints for Snowpark-optimized warehouses, several bug fixes and documentation improvements)
Modin 0.37.0 (bugfixes for Series.json, DataFrame.rename, and eval, and performance improvements for joins with AutoSwitchBackend enabled)
Modin 0.36.0 (bug fix, a performance improvement for query() and eval(), and changes to the testing suite.)
Snowflake VS Code Extension 1.19.1 (bug fix: Setting connection configuration path now accepts directory paths and will transform it into the expected file path)
Streamlit 1.50.0 (Charts: Width option for st.line_chart, Sort option for st.bar_chart, Height option for st.graphviz_chart, Chart column color configuration; Widgets: MultiselectColumn support in st.dataframe and st.data_editor, Default parameter for st.tabs, Border option for st.table, New states for CopyButton (hover, active, focus-visible), Button widgets keyed by identity, Time/date input widgets keyed by identity, Selectbox and multiselect keyed by identity, Sample rate option for st.audio_input, Metric supports decimal values, Slider tick labels on hover; Theming: Theming font source support, Theme main color options, Theme background color options, Theme text color options)

Client, Drivers, Libraries and Connectors Updates

New features:

Snowflake Connector for ServiceNow® V2 5.25.0 (Behavior changes: For records in the sys_created_on or sys_updated_on columns with null values, the connector inserts an update event only when the record has changed since the last ingestion. Previously, the connector inserted an update event to the event log table during each ingestion cycle, regardless of whether the record changed. This behavior could cause the event log table to grow indefinitely, even if no changes were found in the table)
Go Snowflake Driver 1.17.0 (Added support for Go 1.25, dropped support for Go 1.22, added ability to configure OCSP for each individual connection, added DECFLOAT support. See the gosnowflake documentation for details, added proxy options to connection parameters, added client_session_keep_alive_heartbeat_frequency connection paramameter, added support for multi-part downloads for S3, Azure and Google Cloud, added Config.singleAuthenticationPrompt to control authentication flow. When true, only one authentication is performed at a time, allowing for manual interactions such as MFA or OAuth. Default is true)
Node.js 2.3.0 (Added certificate revocation list (CRL) validation support. For configuration options, see Certificate revocation list (CRL) options)
Snowflake CLI 3.12.0 (Added the !edit command to the snow sql command to support external editors, added the — partial option to the snow logs command to support partial, case-insensitive matching of log messages, improved parsing !source with trailing comments, upgraded to typer=0.17.3 to improve the display of help messages, improved output handling with streaming queries in the snow sql command)
Snowflake Connector for Python 3.17.4 (Added support for allowing intermediate certificates from the trust store to act as root certificates, updated bundled urllib3 to version v2.5.0, updated bundled requests to version v2.32.5, dropped support for OpenSSL versions older than 1.1.1)
Snowflake Python API 1.8.0 (Added support for proxy configuration. You can provide proxy settings by using the HTTPS_PROXY environment variable)
Snowpark Library for Python 1.39.0 (New features: downgraded to level logging.DEBUG — 1 the log message saying that the Snowpark DataFrame reference of an internal DataFrameReference object has changed, eliminate duplicate parameter check queries for casing status when retrieving the session, retrieve DataFrame row counts through object metadata to avoid a COUNT(*) query (performance), added support for applying the Snowflake Cortex function Complete, introduce faster pandas: Improved performance by deferring row position computation, the following operations are currently supported and can benefit from the optimization: read_snowflake, repr, loc, reset_index, merge, and binary operations, if a lazy object (e.g., DataFrame or Series) depends on a mix of supported and unsupported operations, the optimization will not be used, updated the error message for when Snowpark pandas is referenced within apply, added a session parameter dummy_row_pos_optimization_enabled to enable/disable dummy row position optimization in faster pandas, dependency updates: Updated the supported modin versions to >=0.35.0 and <0.37.0 (was previously >= 0.34.0 and <0.36.0); Snowpark local testing updates: added support to allow patching functions.ai_complete)
Snowpark Library for Python 1.38.0 (New features Added support for the following AI-powered functions in functions.py: ai_extract, ai_parse_document, ai_transcribe, added time travel support for querying historical data: Session.table() now supports time travel parameters: time_travel_mode,statement, offset, timestamp, timestamp_type, stream, dataFrameReader.table() supports the same time travel parameters as direct arguments, dataFrameReader supports time travel via option chaining (e.g., session.read.option(“time_travel_mode”, “at”).option(“offset”, -60).table(“my_table”)), added support for specifying the following parameters to DataFrameWriter.copy_into_location for validation and writing data to external locations: validation_mode, storage_integration, credentials, encryption, added support for Session.directory and Session.read.directory to retrieve the list of all files on a stage with metadata, added support for DataFrameReader.jdbc(Private Preview) that allows the JDBC driver to ingest external data sources, added support for FileOperation.copy_files to copy files from a source location to an output stage, added support for the following scalar functions in functions.py: all_user_names, bitand, bitand_agg, bitor, bitor_agg, bitxor, bitxor_agg, current_account_name, current_client, current_ip_address, current_role_type, current_organization_name, current_organization_user, current_secondary_roles, current_transaction, getbit; Snowpark pandas API Updates: completed support for the more functions on the “Pandas” and “Ray” backends, added support for Index.get_level_values())
Snowflake ML 1.15.0 (Behavior changes: Model Registry behavior changes: drop support for deprecated conversational task type for Huggingface models. This task type has been deprecated by HuggingFace for some time and is due for removal from their API in the imminent future)
Snowflake ML 1.14.0 (New ML Jobs features: The additional_payloads argument of the MLJob.submit_* methods has been renamed imports to better reflect its purpose. additional_payloads has been deprecated and will be removed in a future release)
Snowflake ML 1.13.0 (New Model Registry features: you can now log a HuggingFace model without having to load the model in memory using huggingface_pipeline.HuggingFacePipelineModel. Requires the huggingface_hub package. To disable downloading from the HuggingFace repository, pall download_snapshot=False when instantiating huggingface_pipeline.HuggingFacePipelineModel, you can now use XGBoost’s enable_categorical=True models to with pandas DataFrames, when listing services, the PrivateLink inference endpoint in shown in the ModelVersion list.)
Snowflake ML 1.12.0 (New Model Registry features: log text-generation models with signatures compatible with OpenAI chat completion compatible signatur, New Model Monitoring features: segment columns to enable filtered analysis, specified in the segment_columns field in the model monitor source options. Segment columns must exist in the source table and be of string type. add_segment_column and drop_segment_column methods are provided to add or remove segment columns in existing model monitors)
Snowpipe Streaming SDK 1.0.0 (released the SDK to support the general availability of Snowpipe Streaming with high performance architecture for AWS deployments, performance and stability improvements)

Bug fixes:

Snowflake Connector for ServiceNow® V2 5.25.2 (Fixed an issue that caused the schema table to be created incorrectly during state import, fixed an issue in filtered reload mode where some state events could be saved in the wrong order, which could lead to missed updates, improved handling of large responses from ServiceNow, added more detailed logging for ServiceNow response properties)
Snowflake Connector for ServiceNow® V2 5.25.1 (Fixed an issue that could cause the metadata tables to be ingested incorrectly during reload when the connector is globally configured to fetch display values. As a result of this issue, flattened views were not created for some tables. If this issue occurred, the following metadata tables had to be reloaded: sys_dictionary, sys_db_object, sys_glide_object)
Snowflake Connector for ServiceNow® V2 5.25.0 (increased the range of page sizes that the connector tries during filtered ingestion. When fetching data, the connector should now be more resilient to timeout errors that come from the ServiceNow® API, fixed the internal cleanup job to retain internal connector information that is needed to perform ingestion. Previously, when this information was removed, it could cause ingestion failures, fixed an error during the creation of flattened views. This error was caused by a missing column in the internal connector table)
Go Snowflake Driver 1.17.0 (Fixed missing DisableTelemetry option in Config, fixed multistatements in large result sets, fixed unnecessary retries when a context is cancelled, fixed a regression in loading TOML connection files, fixed race conditions in stage downloads)
Node.js 2.3.0 (Improved debug logs when dowloading query result chunks, fixed missing error handling in getResultsFromQueryId(), fixed invalid transformation of null values to “” when using stage binds, extended typing of Bind)
Snowflake CLI 3.12.0 (Fixed crashes with older x86_64 Intel CPUs., fixed the ! commands in snow sql commands so they no longer require a trailing ; for evaluation, fixed using ctx.var in snow sql with Jinja templating, fixed issues when pasting content with trailing new lines, fixed an issue with snow snowpark deploy failing on duplicated packages, fixed an issue causing a snow spcs logs IndexOutOfRange error)
Snowflake Connector for Python 3.17.3 (Enhanced configuration file permission warning messages: Improved warning messages for readable permission issues to include clear instructions on how to skip warnings using the SF_SKIP_WARNING_FOR_READ_PERMISSIONS_ON_CONFIG_FILE environment variable, fixed the bug with staging pandas dataframes on AWS — the regional endpoint is used when required: This fix addresses the issue with the create_dataframe call on Snowpark)
Snowpark Library for Python 1.39.1 (Added an experimental fix for a bug in schema query generation that could cause invalid SQL to be genrated when using nested structured types)
Snowpark Library for Python 1.39.0 ( Fixed an issue with drop_duplicates where the same data source could be read multiple times in the same query but in a different order each time, resulting in missing rows in the final result. The fix ensures that the data source is read only once, fixed a bug with hybrid execution mode where an AssertionError was unexpectedly raised by certain indexing operations)
Snowpark Library for Python 1.38.0 (Bug fixes: fixed the _repr_ of TimestampType to match the actual subtype it represents, fixed a bug in DataFrameReader.dbapi that UDTF ingestion does not work in stored procedures, fixed a bug in schema inference that caused incorrect stage prefixes to be used; Improvements: enhanced error handling in DataFrameReader.dbapi thread-based ingestion to prevent unnecessary operations, which improves resource efficiency, bumped cloudpickle dependency to also support cloudpickle==3.1.1 in addition to previous versions, improved DataFrameReader.dbapi (Public Preview) ingestion performance for PostgreSQL and MySQL by using a server-side cursor to fetch data; Improvements: set the default transfer limit in hybrid execution for data leaving Snowflake to 100k, which can be overridden with the SnowflakePandasTransferThreshold environment variable. This configuration is appropriate for scenarios with two available engines, “pandas” and “Snowflake,” on relational workloads, improved the import error message by adding — upgrade to pip install “snowflake-snowpark-python[modin]” in the message, reduced the telemetry messages from the modin client by pre-aggregating into five-second windows and only keeping a narrow band of metrics that are useful for tracking hybrid execution and native pandas performance, set the initial row count only when hybrid execution is enabled, which reduces the number of queries issued for many workloads, added a new test parameter for integration tests to enable hybrid execution; Bug fixes: raised NotImplementedError instead of AttributeError on attempting to call Snowflake extension functions/methods to_dynamic_table(), cache_result(), to_view(), create_or_replace_dynamic_table(), and create_or_replace_view() on DataFrames or series using the pandas or ray backends)
Snowflake ML 1.12.0 (Model Registry bug fixes: fixed an issue where the string representation of dictionary-type output columns was being incorrectly created during structured output deserialization, losing the original data type, fixed an inference server performance issue for wide (500+ features) and JSON inputs)
SQLAlchemy 1.7.7 (fixed an issue that threw an exception for structured type columns dropped while collecting metadata)

Conclusion

Snowflake’s September 2025 release emphasizes three main themes: real-time processing, AI-native features, and governed interoperability. Snowpipe Streaming (GA) provides reliable low-latency ingestion at scale through a throughput-based approach. The FILE data type (GA) enables multimodal AISQL patterns without needing to duplicate blobs. Snowsight Workspaces (GA) and new Query Insights enhance daily development capabilities. In AI, DocumentAI models are now part of the Registry, Cortex Agents feature a REST API and integrate with Teams/Copilot, and AISQL sees improvements like faster (AI_FILTER), more cost-effective (AI_COUNT_TOKENS), and more versatile features (AI_TRANSLATE GA, dynamic-table support). Governance enhancements include automatic view classification, Snowsight Data Quality, MFA one-time codes, task lineage, and External OAuth for Open Catalog.

The platform layer now includes OpenFlow snowflake native deployments (Preview), Native App controls such as automated grants and feature policies, SPCS on GCP (GA), and broad improvements to Iceberg like partitioned writes, position deletes, tunable compaction, and catalog integration updates. SQL performance and usability see tangible improvements with features like SYS_CONTEXT, read-consistency modes, bind-value introspection, time-series RESAMPLE, and adaptive workload distribution. Client/driver and OSS updates also ensure the ecosystem stays current. Overall, Snowflake is streamlining the process from data ingestion to insights, integrating AI into core workflows, and enhancing enterprise controls. When planning roadmaps, focus on streaming pipelines, AI-assisted transformations (using dynamic tables), and Iceberg-compatible designs — these features are ready for deployment.

Enjoy the reading.

I am Augusto Rosa, a Snowflake Data Superhero and Snowflake SME. I am also the Head of Data, Cloud, & Security Architecture at Archetype Consulting. You can follow me on LinkedIn.

Subscribe to my Medium blog https://blog.augustorosa.com for the most interesting Data Engineering and Snowflake news.

DEV Community